Control of the Quadcopter Position Using Visual Feedback

The method using a visual feedback is proposed in order to control the position of the unmanned aerial vehicles (UAV). The position estimator is based on a Viola-Jones detector, which was prepared in order to make continuous recognition of the quadcopter presence in the image captured by the camera. Typical methods of altitude control, which are based on barometric sensor or radar altimeter, are proper in range of meters or hundreds of meters. However, in a case of range of centimeters or single meters those methods may fail. Presented detector has possibility to find the object in the captured image in real time. In order to develop properly working detector a "learning" procedure of the detector was performed. Then static tests were performed in order to examine the detector's performance. After that a carried out dynamic tests indicated that a visual feedback can be used to find the quadcopter position by processing the video stream. Presented method using visual feedback allows controlling the altitude in range of centimeters or decimeters, depending of setup of the system. At this stage the quadcopter control axes are limited to altitude (Y-axis) and horizontal shift (X-axis). The control algorithm was written in Python 3 using the opencv library. Presented results were obtained experimentally.